SECOND DRAFT: COMMENTS WELCOME
Reading, Writing, and Speech Act Theory:
Prolegomena to any Future Logic of Fiction
Fictional discourse poses a serious problem for contemporary analytic philosophy of language. Since the publication of Naming and Necessity, the overwhelmingly dominant view has been that the meaning of a proper name is simply its referent. This thesis, however, brings with it a whole host of problems. One particularly thorny difficulty is that of negative existentials, sentences of the form ‘N does not exist’ (where ‘N’ is a proper name). Intuitively, some such sentences are true, but the direct reference theory seems to imply that they must be either false or meaningless. After all, if the meaning of a name is just its referent, then a sentence such as ‘Mary does not exist’ is meaningless unless there exists a person Mary who is the referent of the name ‘Mary’. But if such a person exists, then the sentence denying her existence must be false.
Now it might be suggested that negative existentials involve a rather rare and specialized use of names, for which an alternate analysis might well be appropriate. One might, for example, have independent reasons for thinking that existence is not a predicate. Even if this is right, however, the analysis of fictional discourse poses difficulties for the direct reference theory that cannot be so easily avoided. The trouble is that fictional names—‘Sherlock Holmes’ and ‘Bilbo Baggins’, for example—lack referents; fictional characters do not exist. As a result, a simple application of the direct reference view to fictional discourse is entirely untenable—this would entail that claims regarding fictional characters are not only never true, but always meaningless.
The tendency in the philosophy of language has been to treat fictional discourse as a peripheral case, posing little threat to the direct reference view. And, as such, the fact that the analysis of fiction seems to run into intractable difficulties has been downplayed, and the fact the analyses on offer tend to be ad hoc has been tolerated. Typical strategies include treating fictional names as abbreviations for definite descriptions, supposing that fictional names occur within the scope of tacit fictive operators of various sorts, or taking them to directly refer to non-existent objects. But descriptivism regarding fictional names runs into some of the same difficulties that prompted its rejection as a theory of ordinary proper names. Holmes, like Aristotle, might not have done any of the things commonly attributed to him. Appeals to fictive operators, even if well-motivated, are ill-equipped to handle “transfictive” uses of fictional names, that is, uses to make claims about the relations between characters and things external to the specific works in which they occur.  And Meinongianism is a strategy bred of desperation.
But fiction talk is hardly peripheral. The use of fictional names is extremely widespread; water cooler conversations about the previous evening’s episode of “Seinfeld,” for example, or perhaps even “Buffy the Vampire Slayer” are ubiquitous. And since the range of claims made about the characters that occur in fictional works is at least as wide as the range made regarding actual people, any strategy grounded in the anomalousness of fictional predicates is hopeless. Moreover, we think that such claims are often true. As a result, unless fictional discourse can be incorporated in a principled way into a systematic semantic framework with the direct reference theory at its core, the direct reference view will have to be rejected. At least if it is to be understood as offering genuine insight into the nature and functioning of language.
I: Speech Act Theory and the Logic of Fiction
There are a number of importantly different types of fictional discourse that any adequate semantic theory of natural language needs to account for. “Fictive discourse” consists of those sentences that are the products of authorial activity and in part constitute fictional texts. Examples include the following:
“Metafictive discourse” consists of sentences used to make claims describing the events and characters that occur in fictional stories. The following are characteristic:
As above, “transfictive discourse” consists sentences used to make claims about the characters and events that occur in works of fiction, but the claims made concern their relations to things external to the specific works in which they occur. Consider the following:
For present purposes, my focus will be only the former two categories.
Fictive discourse is the product of authorial creative activity, whereas metafictive discourse falls paradigmatically within the purview of reader/appreciators. As a result, a natural way to approach the semantics of fiction is by means of an analysis of the relation between authors and readers. Recently, discussions of the logic of fiction have been dominated by what might be called the “communicative model” of this relation. The fiction producing activity of an author is taken to be the performance of a speech act of a certain sort, with certain characteristic intended effects upon (suitably sensitive) readers. And the analysis of such things as the meaning of fictional discourse and the psychological reactions of readers to fiction typically proceeds in terms of the type of speech act the author is regarded as performing.
Searle, for example, characterizes the author’s speech act as one of pretending to make an assertion. An analysis of the act of asserting might go something as follows:
a speaker S asserts that P by means of her utterance U iff she intends that (i) the audience will come to believe that P by means of (ii) their recognition that S intends them to believe that P, where (iii) the recognition of this intention is achieved by the audience’s recognition (that U means that P and) that S means that P by U. So, for example, Pierre asserts that Basil Rathbone was an actor by means of his utterance “Basil Rathbone was an actor” just in case Pierre intends that Marie (his audience) will come to believe that Basil Rathbone was an actor by means of her recognition that Pierre intends that she believe that Basil Rathbone was an actor by means of her recognition that Pierre means that Basil Rathbone was an actor by his utterance “Basil Rathbone was an actor”. On Searle’s view, authors pretend to perform actions of this sort. And they do this by means of actually performing utterance acts of the same sort as those performed when making an assertion. The difference is that when pretending to make assertions, authors intend to invoke the “horizontal” conventions of fictional discourse, conventions which suspend commitments ordinarily required for felicitous asserting. For example, assertion but not pretended assertion is governed by the “sincerity rule”, a rule requiring the speaker’s commitment to the truth of the expressed proposition.
Currie, in contrast, argues that authors perform a uniquely fictive speech act, an act of “storytelling.” Currie’s analysis of this fictive illocution goes roughly as follows: a speaker S storytells that P by means of her utterance U iff she intends that (i) the audience will come to make-believe that P by means of (ii) their recognition that S intends them to make-believe that P, where (iii) the recognition of this intention is achieved by the audience’s recognition (that U means that P and) that S means that P by U. So for example, Marie storytells that it was a dark and stormy night by means of her utterance “It was a dark and stormy night” just in case Marie intends that Pierre (her audience) will come to make-believe that it was a dark and stormy night by means of his recognition that Marie intends that he make-believe that it was a dark and stormy night by means of his recognition that Marie means that it was a dark and stormy night by her utterance “It was a dark and stormy night.”
I do not intend to adjudicate the dispute between Searle and Currie here. In fact, in my view the speech act model itself of authorial activity is misconceived. More on this below. For present purposes, what I want to focus on is the extent to which these each of these accounts yields promising results for the semantic analysis of fictional discourse. It is worth noting that on Searle’s account, the actions of authors are of central importance, while on Currie’s view, it is the intended effects of these actions of readers that are central.
II: Authorial Pretense
On Searle’s view, authors, when engaging in fictive discourse, pretend to make assertions. And when they produce sentences that contain fictional names, part of this pretense involves pretending to refer to real individuals. Readers and critics, when engaging in metafictive discourse, make genuine assertions. And when they produce sentences containing fictional names, they really refer to fictional characters. Searle says, “I did not pretend to refer to a real Sherlock Holmes; I really referred to the fictional Sherlock Holmes.” Moreover the referents in questions are creations of authors, which authors create by means of their acts of pretended reference: “[by] pretending to refer to people and recount events about them, the author creates fictional characters and events.”
The trouble with Searle’s view is that it is so underdeveloped it is hard to even know how to evaluate it. In lieu of an account of exactly what sorts of fictional characters are poorly placed to discern either (i) whether or not they are the sorts of things that could be created by means of author’s acts of pretense or (ii) whether they are the sorts of things readers do (or could) refer to in their talk about fiction. Searle does, however, offer the following remark: “[it] is the pretended reference which creates the fictional character and the shared pretense which enables us to talk about the character.” The idea seems to be that readers are able to make genuine assertions about fictional characters only when engaged in some sort of “shared pretense”. Perhaps the author’s pretense causes the (suitably sensitive?) reader to pretend, or perhaps imagine, the people the author pretends to refer to engaging in events the author pretends to recount. And perhaps readers so engaged who make genuine assertions using fictional names refer to elements of their imaginings. Perhaps, but even if Searle has something like this in mind, serious worries remain. First, it is far from clear that we do ordinarily engage in any sort of pretense when we make assertions about fiction. Second, an account of the sense of pretense at issue needs to be forthcoming. After all, it is fairly clear that this passive sense of pretense required of readers is fundamentally different from the more active sense Searle relies on in his account of the speech acts performed by authors. Third, an account of the sense in which the pretense is shared needs to be explained. Unless it is shared in a substantial sense, Searle’s view is at risk of implying that fictional names differ in meaning in the mouths of different speakers. Finally, Searle’s account of authorial pretense is at best an ad hoc hodgepodge. Many works of fiction include reference to real people and events. And Searle claims that when authors pretend to make assertions regarding such things, they actually refer to them: “[in] the case of realistic or naturalistic fiction, the author will refer to real places and events…” So, authorial pretense involves actual reference except where it could not. The central problem facing Searle is that even if the pretended assertion theory of authorial speech acts is accurate, an independent account of the proposition the author pretends to assert needs to be provided. And there is no good grounds for Searle’s expectation that pretense can do double duty here.
David Lewis, like Searle, advocates the pretense account of authorial speech acts. But unlike Searle, he recognizes that an independent account of what the author pretends to assert needs to be provided. Strictly speaking, what Lewis offers is an account of the truth conditions of sentences of metafictive discourse rather than of fictive discourse, but an account of the latter can easily be recovered from Lewis’ account of the former. Lewis’ central idea is that since authors pretend to assert the sentences they produce, an account of the propositional content of their utterances is to be found by considering possible circumstances in which these very same sentences are actually asserted. Lewis proceeds by treating fictional discourse as if it occurs within the scope of tacit fictive operators, and gives the following analysis of such operators:
A sentence of the form “In fiction f, S” is true iff “S” is true at every world where f is told as known fact rather than fiction.
So, for example, on Lewis’ view, “Bilbo Baggins left the Shire after his eleventy-first birthday party” is taken to be shorthand for “In The Lord of the Rings, Bilbo Baggins left the Shire after his eleventy-first birthday party.” And this latter sentence is true just in case “Bilbo Baggins left the Shire after his eleventy-first birthday party” is true in every world in which The Lord of the Rings is told as known fact.
An analysis of the right-hand side of this bi-conditional is far from trivial, however. First, an account needs to be given of the conditions under which a fiction is told as known fact at a given world. A necessary but not sufficient condition, according to Lewis, is that there occurs an act (or series of acts) of assertion in which there is a word-for-word, meaning-for-meaning match with the words produced by the author in her act of pretense. In addition it is required that the non-actual act of assertion and the actual act of pretense be the very same act of storytelling. Presumably, this requires this we restrict our attention to worlds in which the actual author asserts the words of the text, unless one assumes that the agent who performs it is an inessential feature of an action.
Second, an account of the conditions under which a sentence of metafictive discourse is true at a non-actual possible world needs to be provided. And this requires an account of the meanings of fictional names when they occur in such sentences. Lewis says the following:
“the sense of “Sherlock Holmes” as we use it is such that, for any world w, the name denotes at w whichever inhabitant of w it is who there plays the role of Holmes. Part of that role, of course, is to bear the ordinary proper name “Sherlock Holmes”. But that only goes to show that “Sherlock Holmes” is used at w as an ordinary proper name, not that it is so used here.”
Lewis’s view is that the meaning of a (fictional?) name when used in metafictive discourse is distinct from its meaning when used as an ordinary proper name at a world in which the fictional work is told as known fact. When used in the latter way, the meaning of a name is (presumably) its referent. And exactly which individual is the referent of a name on an occasion of use is a matter for a theory of referring to determine. If the causal-historical theory is accurate, it will be the individual dubbed in the naming ceremony that initiated the chain of appropriately causally linked events that culminated in the use of the name in question, or something along these lines. The meaning of the fictional name when used in (actual world) metafictive discourse is the function whose value is the occupant of the relevant fictional role at that world—the person who satisfies the descriptions and performed the deeds described in the story—at worlds in which the work of fiction is told as known fact, and is undefined (i.e., has no denotation) at worlds where this does not occur. Note: in order for Lewis’ view to yield accurate predictions regarding the truth values of sentences of metafictive discourse, the value of the function at a world, w, will have to coincide will have to coincide with the referent of the name when used at w as an ordinary proper name. Restricting our attention to those worlds in which the relevant fictional world is told as known fact in meant to ensure this coincidence. More on this below.
There are a number of well-known difficulties that arise for both descriptivist and possible worlds accounts of fictional discourse, from which Lewis’ view is not immune. For example, certain works of fiction are contradictory and, as such, there are no possible worlds in which they are told as known facts. As a result, all claims about such works will turn out trivially true on Lewis’ analysis. What I want to focus on here, however, are problems that can be more or less directly traced to the pretense model of authorial activity. This model guides Lewis’s theory in the following way: insofar as authors in their productive activities pretend to assert the utterances/inscriptions they produce, the most natural account of their propositional contents when so used is one that invokes their contents when used in genuine assertions. And conforming to this “natural” ideal produces a number of difficulties for Lewis’ ultimate view. First, the rules conditions of felicitous assertion incline one in the direction of a view wherein the work of fiction is told as known fact. And this requires not only that the speaker both believe and have evidence for the propositions she expresses, but it also places both the author and her audience (if the assertion is to have a point) in the very world in which the events depicted in the fictional work take place. But these presumptions run afoul of facts regarding actual fictional texts. The events depicted in a fictional work are often incompatible with the narrator’s belief or knowledge of them (or even her existence in the world). Consider, for example the following: “She was innocent, but she took her secret to the grave. Nobody would ever know.” And fictional narrators often lie or have false beliefs regarding the events they describe. Lewis attempts to handle this latter problem by suggesting that in such cases we consider worlds in which the fictional work is told not as known fact, but as a lie or as the expression of false beliefs. The trouble with this idea is that given the wide range of possible worlds in which the fictional work is so told, what goes on in such worlds will prove an unreliable guide to truth in such works.
Second, requiring that a work of fiction be told as known fact at a world does not suffice to guarantee that there will be any occupants of the intuitive fictional roles at all at that world. Consider the following pair of sentences: “Peter is a professor. He got his doctorate at UNC.” The pronoun ‘he’ could be used here either anaphorically to refer to Peter or demonstratively to refer to someone else. And if used in the latter way, the pair of sentences could be told as known fact at a world at which there is no one who is both a professor and who got his doctorate at North Carolina. There are two strategies Lewis might deploy to handle this difficulty. First, he might individuate fictional roles in terms of the set of worlds in which the fictional work is told as known fact. The trouble with this suggestion is that the resulting conception of fictional roles would be too thin. The roles would consist of long disjunctions corresponding to the various networks of coreferential relations that could hold among the singular referring expressions occurring in the fictional work in the various worlds in which it is told as known fact. And this would render many (most?) of the intuitively true (non-disjunctive) sentences of metafictive discourse false. Second, Lewis might restrict the relevant class of worlds to those told-as-known-fact-worlds in the singular terms were produced with appropriate coreferential intentions. Now it might be hoped that this can be achieved by considering only those worlds in which the (non-actual) act of asserting the fictional work is numerically identical to the author’s (actual) act of pretended assertion. Whether or not this will suffice depends, of course, on the correct account of the transworld identity of speech acts. And it is far from obvious that the correct theory must require the sameness of coreferential intentions.
And third, even if this problem can be solved, unless the fictional text contains an explicit uniqueness clause, included in relevant subclass of told-as-known-fact-worlds will be those in which the fictional role is occupied by multiple individuals but in which the use of the fictional name in question refers uniquely to only one of them. One might attempt to avoid this problem by analyzing actual uses of fictional names as universal quantifications over fictional roles. For example, one might analyze “[It is Sherlock Holmes fictional that] Holmes was a detective” as “[It is Sherlock Holmes fictional that] all occupants of the Holmes-role were detectives” or something along these lines. Suppose, however, that the sentence “Mary told only one person in the whole world her deep dark secret: either Dan or Margaret” occurred in a fictional work (“The Secrets of Mary”, let’s call it) that was told as known fact at a world w at which there are two occupants of the Dan-role. Moreover, suppose that it is not part of the Dan-role that he either was or was not Mary’s confessor, but the speaker knew the sentence to be true because she knew that Mary told one occupant of the Dan-role her secret. On the analysis under consideration, the corresponding sentence about fiction “[It is “The Secrets of Mary” fictional that] Mary told only one person in the whole world her deep dark secret: either Dan or Margaret” would be false at w, because no occupant of the Margaret role was Mary’s confessor there nor were all occupants of the Dan-role.
III: Readers’ Make-Believe
Kendall Walton models his theory of fiction on children’s games of make-believe. Games of make-believe are rule-governed activities. Such rules prescribe that certain propositions be imagined; these are the fictional truths of the game in question. A prop in a game of make-believe is an object that in virtue of the rules of the game generates fictional truths. Suppose, for example, a rule of a game is that if there is a dog in a certain place, we are to imagine there is a lion there. A dog in a neighbour’s yard would be a prop in such a game and it would generate the fictional truth that there is a lion in the yard. Fictional works are objects whose function is to serve as props in certain games of make-believe, where “[a] thing may be said to have the function of serving a certain purpose, regardless of the intentions of its maker, if things of that kind are typically or normally meant by their makers to serve that purpose.”  The games in which it is the function of a work to serve as a prop are the games that are authorized for it.
Walton uses this apparatus in order to paraphrase ordinary talk about fiction as follows:
A sentence of the form “In fiction f, S” is true iff fiction f is such that one who engages in pretense of kind K in a game authorized for it makes it fictional of herself that she speaks truly.
So, for example, “[In The Lord of the Rings,] Bilbo Baggins left the Shire after his eleventy-first birthday party” would, on Walton’s analysis, be paraphrased as “The Lord of the Rings is such that one who engages in pretense of kind K in a game authorized for it makes it fictional of herself that she speaks truly.” The crucial question here, of course, concerns what constitutes a pretense of the requisite kind. On Walton’s view, the speaker who pretends to assert the sentence about fiction under analysis (while at the same time asserting the Waltonian paraphrase) displays an example of the type of pretense in question. And the speaker who produces the sentence about fiction, in a way that is not in the “spirit of pretense”, does not display an example of the requisite type but goes through the motions of so doing.
The question that remains concerns in what respect an act of pretense must be similar to the displayed example (or the sentence produced going through the motions of pretense) in order for it to be of kind K. Walton’s answer to this question depends on whether or not the sentence under analysis contains any fictional names. If it does not, an act of pretending to assert is of the same kind as the displayed example just in case the sentence used in the former in fact expresses the same proposition as the sentence used in the latter. If it does contain a fictional name, however, things become more complex. The reason is that, on Walton’s view, such sentences express no propositions. Walton’s move here is to suppose it can be fictionally true that two distinct speech acts are assertions of a single proposition despite the fact that the utterance/inscription produced in neither act of pretense expresses a proposition. And when this condition is met, the two acts of pretense are of the same kind for the purposes of Walton’s analysis. Suppose, for example, Peter pretended to assert the sentence “Bilbo Baggins left the Shire after his eleventy-first birthday party”. Peter would at the same time be asserting the Waltonian paraphrase under consideration. In such circumstances, kind K would consist of the class of acts of pretense each of whose members, a, is such that it is fictionally true that a and Peter’s act are assertions of the same proposition. And Peter’s assertion would be true only if the rules governing authorized The Lord of the Rings games prescribe imagining that each member of K is a true assertion.
In my view, however, Walton’s analysis falls prey to the following dilemma: unless the rather elaborate apparatus he deploys proves ultimately to be superfluous, his theory will remain shrouded in mystery. The central datum I wish to invoke here is the systematic network of meaning relations that hold between the sentences that occur in fictional texts and the true sentences about such fictions. Typically, if a declarative sentence, “S”, occurs in a fictional work, f, then the sentence “In fiction f, S” is true. Moreover, for any sentence “T”, such that “T” means the same as “S”, the sentence “In fiction f, T” will be true as well. On Walton’s view, a sentence about fiction is true only if the rules governing games authorized for the fiction work in question prescribe imagining that pretended assertions of it are (actual assertions and) true. Now the aforementioned systematicity can be explained on Walton’s view if it is supposed that there is a rule governing all games authorized for all fictional works to the effect that one ought to imagine that all pretended assertions of sentences meaning equivalent to declarative sentences occurring in the text are true. But if this is right, an equally materially adequate but less complex account of fictional truth can be given in terms of the meaning relations themselves. However, if there is no such rule, the systematicity in question becomes mysterious on Walton’s view.
It might be objected, however, that meaning relations cannot be invoked when discussing sentences that utilize fictional names. After all, they do not express propositions at all. The trouble with this line of response is that fictional names, despite lacking actual referents, do stand in intuitive coreferential relations. And truths about fiction systematically reflect these coreferential relations. Now providing an account of “coreference” in fiction without actual reference is, of course, a thorny business. But again, either Walton will have to formulate his prescriptive rules (in this case for unofficial extended games) in terms of some such theory or leave the systematicities mysterious.
As we have seen, even if the communicative model adequately characterized the relation between authors and readers, an independent account of the semantics of fictional discourse would still be required. Both authorial pretense and reader make-believe are actions directed towards propositional contents. And as result, any adequate analysis of them presupposes an account of their propositional objects. Searle and Walton run into trouble by trying to generate an account of the propositional objects of such acts from an account of the acts themselves.
But the communicative model itself is misconceived. It reflects a troubling equivocation between the activities of authors and the activities of performers. Consider, for example, the following comment from Lewis:
“Rather, a fiction is a story told by a storyteller on a particular occasion. He may tell his tales around the campfire or he may type a manuscript and send it to his publisher, but in either case there is an act of storytelling.”
Contra Lewis, however, typing a manuscript and telling a story around a campfire are distinct sorts of activities. They are both (partially) constituted by the production of sequences of linguistic tokens, but only the former involves the performance of a speech act. In order to perform a speech act, you need to do more than produce a linguistic token; you need to do something with it, or use it in a certain way. The distinction between writing a story and telling it is not the distinction between engaging in written and spoken language use. Fictional texts can be produced by either typing or dictating. And written tokens can be used in the performance of speech acts. One might, for example, request that someone be quiet by gesturing to her with a placard on which is written, “Please, be quiet.” The central point is that there is no characteristic behaviour authors engage in using the texts they produce; and rarely if ever can the behaviours they in fact engage in be plausibly construed as language use. Some authors send their texts to publishers; others publish their texts themselves. Some authors give them to friends; others tuck them away in their underwear drawers. And while these various actions are purposeful, the intended effects—that the work be published, or read—are not primarily communicative. Nor can it plausibly be maintained that these are perlocutionary effects intended to be achieved by means of the successful performance illocutionary acts of various sorts involving the constitutive sentences of the text. Of course, an author might hope to achieve certain further effects by means of publishing her work similar to those achieved by means of the performance of genuine speech acts. An author of a work of non-fiction, for example, might hope to instill certain beliefs in the reading public. But the author’s behaviour does not thereby count as an act of assertion, because the causal pathway from authorial activity to intended effect is too arbitrary. Putting a manuscript in the mail, or perhaps, signing a publication contract, is simply not in any sense a conventional procedure for achieving desired effects on the psychological states of others.
Searle, in his discussion of drama at least, distinguishes between writing and performance: performers (in plays) pretend to make assertions; playwrights provide them with instructions for their pretense. But even if this is an accurate account of the activity of playwrights, it will not do as an account of authorial activity more generally. Novels typically are not performed and are not characteristically written for said purpose. Moreover, even if they are performed, this performance is inessential to reader understanding of the text.
Authorial activity is better understood in terms of a craft-making model. The goal of this activity is the production of an artifact of a certain sort, in particular one that conforms to the regulative ideals of novel writing. These ideals presumably have been shaped by the reactions of readers to written works of various sorts. And authors, by producing artifacts that more or less conform to these ideals, may well hope to achieve various effects on their readers. But the central goal of this activity is the production of a novel. Now it is true that producing an artifact of this sort involves producing an extended sequence of sentences. These sentences, however, are displayed tokens lacking illocutionary force; the author no more performs a speech through their production than does the traffic-sign maker command us to stop. Moreover, viewing texts in this way has certain advantages. In particular, it enables us to regard the sentences that occur in fictional works—including those that contain fictional names—as meaningful without being forced to suppose that they have determinate propositional contents. Now of course authors typically have conceptions of individuals and events that their novels are meant to describe, and they may have rich networks of intentions regarding which terms denote which individuals and events in these conceptions. But this does not suffice to show that the sentences they produce express propositions regarding such mental entities.
It might be argued that even if the communication model is rejected, an analysis of fiction in terms of readers’ make-believe is still appropriate. After all, a central datum of fictional discourse is that the paradigmatic reader attitude towards the sentences in fictional texts is not that of belief (or even disbelief) but some other attitude. We might even distinguish this “fictional attitude” from belief in terms of its functional role. Now insofar as ‘make-believe’ is a term of art for whatever this fictional attitude is, I have no objection to its use, although it is not particularly helpful. But insofar as it denotes participation in rule-governed imaginative games, as Walton suggests, it is wrongheaded for (at least) two reasons. First, it simply does not capture the phenomenology of the experience of reading. And second, Waltonian make-believe arguably occurs only when we lose focus on the text and drift off into reverie, rather than when we are engaged readers of fiction.
A better model, in my view, is one that takes the fictional attitude to consist of reader assent to the sentences of fiction, rather than belief. I do not currently have a full-blown theory of assent, but there are a few comments I would like to make in favour of the assent model. First, there is a tolerably clear, pre-theoretical notion of assent available—namely that of accepting sentences for the purpose of hypothetical reasoning—which is not obviously at odds with the experience of reading. What I have in mind is what one does when, for example, one is asked to suppose that rabbits can talk, and then consider what they might say. And this attitude clearly falls short of believing (the proposition expressed by) the sentence in question. Second, acceptance, unlike belief, brings with it no commitment on the part of the acceptor to the existence of the denotata of the subsentential expressions of the accepted sentences. An engaged reader, especially of Russian novels, might well be concerned to figure out when the fictional names she encounters corefer, but this brings with it no commitment to the existence of the referents of such names.
V: Sketch of a Logic of Fiction
In this section, I want to provide a preliminary sketch of the approach to the logic of fiction I will be developing down the road. There are some interesting similarities between the picture I wish to defend and Currie’s view, and so a brief discussion of this latter view seems in order. Central to Currie’s theory is the notion of a fictional author, “that fictional character within our make-believe whom we take to be telling us the story as known fact.” And he uses this notion to give the following account of truth in fiction:
A sentence of the form “In fiction f, S” is true iff it is reasonable for the informed reader to infer that the fictional author of f believes that “S” is true.
Currie’s view has a number of nice features. First, in a work of fiction a disjunction can be true without either of its disjuncts being true. For example, it is presumably true that either Holmes has a mole on his left shoulder or he does not. Holmes is a person, after all, and disjunctions of that sort are true of all persons. But it is not true that he does and it is not true that he does not: the Holmes corpus simply does not determine the truth of one or the other disjunct. As a result, an analysis in terms of belief is apt because people can believe disjunctions without believing either of their disjuncts. Second, by offering an analysis in terms of the beliefs of the fictional narrator rather than in terms of features of the possible world in which she narrates, Currie’s view (unlike Lewis’) does not imply that all claims about contradictory works of fiction are trivially true. After all, a person can believe each member of a pair of contradictory propositions without believing everything. And finally, Currie provides an account of those truths about fiction which extend beyond what is explicitly stated in texts that explains both why such an account has proved so difficult to provide and why such truths are often so controversial. Determining what a fictional author believes is not simply a matter of determining the logical implications of what she has said, but is a matter of holistic interpretation governed by the regulative ideal of rationality.
There are, however, a couple of difficulties with Currie’s view. First, fictional narrators (at least explicit ones) are often quite unreliable: sometimes they have false beliefs about the events they relate; sometimes they simply lie. Moreover, sometimes novels have numerous fictional narrators with contradictory beliefs. As a result, any account that identifies fictional truth with the beliefs of a fictional narrator will be inadequate. Currie attempts to solve this problem by distinguishing between an explicit fictional narrator and a fictional author, “one who tells a story he knows to be true by speaking with the voice of one of the (unreliable) characters in the story.” But not only is this move ad hoc, Currie offers little in the way of principled grounds for distinguishing between those texts wherein the narrative voice is that of reliable fictional author and those in which a fictional author speaks with the voice of an unreliable fictional narrator. And second, Currie’s offers a descriptivist account of fictional names, but one which is both unmotivated and oddly disconnected from his account of fictional truth. He simply stipulates that because they lack referents they cannot be genuine proper names.
The view I am going to sketch here has the virtues of Currie’s account while avoiding the difficulties. There are two central features of this view that I wish to delineate at this point. First, like Currie I think it is useful to invoke the notion of a fictional narrator in an account of fictional truth. But instead of analyzing fictional truth in terms of the beliefs of the fictional narrator, on my view it is to be analyzed in terms of what the fictional narrator said. The guiding idea involves treating the fictional text as if it was uttered by a fictional narrator and defining fictional truth in light of this supposition roughly as follows:
A sentence of the form “In fiction f, S” is true if (and only if?) “The fictional narrator of f said that S” is true.
As with belief, it does not follow from the fact that a fictional narrator asserted a disjunction that she asserted either disjunct. Nor does it follow from the fact that she asserted a contradiction that she asserted everything. The advantage of this view is that it is compatible with the possibility of fictional narrators who lie or have false beliefs. Moreover, it fits nicely with the assent model of the reader’s fictive attitude. When one assents to something one has been told without believing it, in reporting the conversation one appropriately makes claims only about what was said and not how things are. Finally, an account of truth in fiction beyond what is explicitly said in the text might be formulated in terms of what the fictional narrator pragmatically imparts by means of what she says.
Second, this approach suggests an account of fictional names according to which they function as genuine proper names despite their failure to have referents. Indirect-discourse contexts are opaque contexts, that is, they are contexts in which co-referential names are not intersubstitutable salva veritate. For example, the truth of “Lois Lane said that Clark Kent can fly” cannot be inferred from the truth of “Lois Lane said that Superman can fly”. A familiar strategy for handling opaque contexts, and the one that I favour, invokes Frege’s reference shift idea. The idea is that in such contexts, the meaning of a proper name is not its ordinary (extensional) referent, but its customary sense. As a result, analyzing talk about fiction in terms of what fictional narrators say facilitates a view according to which fictional names are genuine proper names and meaningful, despite lacking referents.
Finally, let me present a preliminary sketch of the account of indirect discourse I mean to deploy in my theory of fiction. In what follows I will refer to the person regarding whom the indirect discourse ascription is made as the “original speaker” and the person who makes the ascription as the “ascriber.” Suppose the original speaker (OS) utters the sentence, ‘Basil is an actor.’ The meaning of OS’s use of ‘Basil’ is an individual person, in particular, the individual dubbed in the naming ceremony that initiated the chain of appropriately causally linked events that culminated OS’s use of ‘Basil.’ What is important to note is that although OS’s conception of the person she took herself to refer to by means of her use of ‘Basil’ did not (even in part) constitute the meaning of the name, she did conceive of him in some way or other nonetheless. Now suppose that OS’s utterance of ‘Basil is an actor’ prompts the ascriber (A) to say ‘OS said that Basil is an actor.’ (Moreover, suppose that A’s ascription is opaque and not transparent; that is, given the conversational purpose underlying A’s utterance, he meant to draw attention not to the object of OS’s prior talk but to her mode of conception of said object). A’s use of ‘Basil’ is appropriately causally related to OS’s use of the name, as was Os’s use to the prior uses of which it was the culmination. But given the conversation purpose of A’s utterance, the meaning of her use of ‘Basil’ is not the referent of OS’s use, but OS’s conception of said referent.
This model can be applied to fictional discourse as follows. As above, fictive discourse consists of displayed sentential tokens, some of which at least lack propositional content. As a result, the fact that fictional names lack referents poses no difficulties for fictive discourse. The indirect discourse model of metafictive discourse involves treating the sentences of fictive discourse as if they were the utterances of a fictional narrator. The sentences of metafictive discourse are analyzed as indirect discourse reports of the utterances of the fictional narrator. Given that that these reports are opaque, the meanings of fictional names in such reports are (individual) concepts of fictional narrators. Consider once again the following sentence of metafictive discourse: “In The Lord of the Rings, Bilbo Baggins left the Shire after his eleventy-first birthday party.” On the indirect discourse model, this gets analyzed (roughly) as, “The fictional narrator of The Lord of the Rings said that Bilbo Baggins left the Shire after his eleventy-first birthday party.” And the meaning of ‘Bilbo Baggins’ is the fictional narrator of The Lord of the Rings’ concept of Bilbo Baggins. A lot more needs to be said here, but this will have to do for present purposes.
The communicative model of fiction is inadequate and ought to be abandoned. Not only does it fail to provide insight into the semantics of fictional names in its won right, it tends to distort views on this matter. The craft-making model of authorial activity and the assent model of readers’ fictive attitudes together offer better guidance on this issue. And the displayed-token theory of fictive discourse and the indirect discourse theory of metafictive discourse that they suggest both seem promising, at least at a first glance. But we shall see.
 Saul Kripke, Naming and Necessity, Cambridge: Harvard University Press, 1980. Strictly speaking, what Kripke defended was the weaker thesis that names are rigid designators.
 Gregory Currie, The Nature of Fiction, Cambridge: Cambridge University Press, 1990, p.171.
 Currie (1990) p. 146.
 Currie (1990) p. 158.
 John Searle, “The Logical Status of Fictional Discourse,” New literary History 6: 319-332, 1975.
 This account is mean to mirror the account of fictive speech acts given by Currie (1990, p. 31). Searle (1975, p. 322) also offers a number a number of pragmatic rules of felicitous assertion.
 Searle (1975) pp. 326-7.
 This is a variation on the first of the three formulations Currie provides (1990, p. 31).
 Searle (1975), p. 330.
 Searle (1975), p. 331.
 Searle (1975), p. 330.
 Searle (1975), p. 331.
 David Lewis, “Truth in Fiction”, American Philosophical Quarterly 15:37-46, 1978.
 Lewis (1978) p. 41. On this analysis, nothing is true in fiction except what is explicitly stated. Lewis also offers other analyses which take into consideration the commonly held intuition that it is appropriate to read into fiction background assumptions of various sorts.
 Lewis (1978) p. 40. Lewis invokes this second condition in order to handle cases in which one author produces a text word for word identical to one produced previously by another author. Presumably, however, the two texts will yield different fictional truths only if fictional truths generated by a text extend beyond what is explicitly uttered, and vary with (perhaps contextual) facts about the author. On this analysis under consideration this is not the case.
 This strikes me as being as implausible as suggesting that there are possible circumstances in which someone else smiles my very smile.
 Lewis (1978) p. 41.
 Lewis (1978) p. 41. Lewis includes this latter clause to prevent “Sherlock Holmes” from denoting some actual person who was unknown to Doyle and yet in fact occupied the Holmes-role.
 But for the problem of fictional names, one could take the contents of the actual and pretended assertions to coincide.
 See, e.g., Searle (1975) p. 322.
 Lewis (1978) n. 7.
 There might even be worlds in which the fictional work is told as known fact on two separate occasions, where the referent of the fictional name on the first telling is one occupant of the fictional role while its referent on the second telling is another occupant of the fictional role.
 Kendall Walton, Mimesis as Make-Believe, Cambridge: Harvard University Press, 1990. I am focusing on Walton’s account of here because make-believe ends up playing little substantial role in Currie’s account of the semantics of fictional discourse. But it is worth noting that Walton explicitly denies that fiction-making is a speech act of any sort (p. 81).
 Walton (1990) p. 52.
 Walton (1990) p. 400.
 Walton (1990) pp. 402-4.
 Walton (1990) p. 403.
 Walton acknowledges that in order for acts of pretense preformed by distinct people to be of the same kind, things will have to be formulated in terms of unofficial extended games understood to include both actions.
 Apologies to Meinong and his followers.
 Lewis (1978) p. 39.
 Consider, for example, an author whose works get published on reputation alone.
 Searle (1975) p. 328.
 Of course, improvised storytelling is something of a hybrid case.
 One might, of course, take reading to be a private performance for oneself, but this would require fairly strong motivation in order to be made even plausible.
 The ideal for works of non-fiction would presumably include conformity to truth. For most genres of fiction, the ideal would likely be a purely aesthetic one.
 They are meaningful in the same sense that “this is a nice red one” is meaningful when “this” lacks a demonstratum.
 For example, if I believe that someone is drowning, I will be inclined to engage in some form of rescue securing behaviour. But if I adopt the “fictional attitude” towards the sentence ‘Someone is drowning”, I will be inclined to do no such thing.
 This point is due to John Woods.
 Currie (1990) p. 76.
 Currie (1990) p. 80.
 Currie (1990) p. 74. Currie delineates a number of such similarities between belief and fictional truth at this point.
 Currie (1990) pp. 87-8.
 Currie (1990) p. 125.
 Currie (1990) p. 159.
 One worry would be how I might handle a novel with a number of distinct explicit narrators who offer conflicting accounts of events. My suggestion is to continue to take what is true in a fictional narrative to coincide with what the fictional narrator says, but to suppose that certain novels are composed of a series of distinct fictional narratives. If someone insists that what some narrators say is accurate and others inaccurate, I can always capture this notion by appeal to features of the world in which the narrators exist. But in my view, this is a derivative notion of truth in fiction. What is fundamental is the story we are told.
 This might yield the following account of fictional truth:
A sentence of the form “In fiction f, S” is true if and only if “The fictional narrator of f said that S or pragmatically imparted that S” is true.
 Gottlob Frege, “On Sense and Reference”, in The Frege Reader, M. Beaney, ed., Blackwell, 1997.
 Exactly what causal role her conception of the referent played in producing her use of the name, and in producing an appropriate link to prior uses of it, is a matter best left for another time.